Investigating association rules for sentiment classification of Web reviews

Sentiment Classification of web reviews or comments is an important and challenging task in Web Mining and Data Mining due to the increasing social media and e-commerce industry. This paper presents a novel approach using association rules for sentiment classification of web reviews. An optimal classification rule set is generated to abandon the redundant general rule with comparatively lower confidence. In the class label prediction procedure, we proposed a new metric named Maximum Term Weight (MTW) for the evaluation of rules and a multiple metric voting scheme to solve the problem when the covered rules are not adequately confident or not applicable. The final score of a test review depends on the overall contributions of four metrics. Experiments on multiple domain datasets from web site demonstrate that the voting strategy obtains improvements on other rule based algorithms. Another comparison to popular machine learning algorithms also indicates that the proposed method outperforms these strong benchmarks.